The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/metrics_engine.py file executes git show commands via subprocess.run. It incorporates the repo_root path and commit_sha identifiers into the command arguments. While it avoids shell interpolation, executing sub-processes with externally-sourced parameters is a monitored behavior.
[REMOTE_CODE_EXECUTION]: In scripts/metrics_engine.py, the load_module function uses importlib.util to dynamically load and execute modules from other skill directories (e.g., guide-planning, add-subfeature). This runtime execution of scripts from relative paths constitutes dynamic code loading.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted markdown and metadata.
Ingestion points: Processes slice-traceability.md and metadata files in scripts/metrics_engine.py.
Boundary markers: None. Output is produced without clear delimiters or instructions to ignore embedded commands.
Capability inventory: The skill can execute shell commands and write to files.
Sanitization: Content from the file system is not sanitized before being rendered to the agent.

measure-artifacts